Granted Patent B2

US 9,008,422 · App. 13/721,778 · Granted Apr 14, 2015

Abstract

Claims (48)

1. A method of cropping an image based on a content of the image, comprising:

receiving the image at a processor;

identifying image features in the content of the image;

determining horizontal histogram frequencies and vertical histogram frequencies of the identified image features;

determining borders of a document bounding box based on the determined horizontal histogram frequencies and vertical histogram frequencies;

determining a horizontal centroid and a vertical centroid of the horizontal and vertical histogram frequencies, respectively;

resizing the document bounding box based on the determined horizontal and vertical centroids; and

cropping the image according to the document bounding box as said resized.

2. The method of claim 1 , further comprising downscaling the image prior to the cropping the image.

3. The method of claim 2 , further comprising upscaling the document bounding box, wherein cropping the image according to the document bounding box includes cropping the image according to the upscaled document bounding box.

4. The method of claim 1 , wherein identifying image features in the content of the image includes identifying salient elements in the content of the image as image features and ignoring extraneous elements in the content of the image.

5. The method of claim 1 , wherein identifying image features in the content of the image includes determining a gradient of the image and identifying an image feature where a magnitude of the gradient of a portion of the image satisfies a predetermined threshold.

6. The method of claim 1 , wherein determining borders of the document bounding box includes:

determining a left border and a right border of the document bounding box by determining, from left to right of the image and from right to left of the image, respectively, whether a horizontal histogram frequency value exceeds a predetermined threshold; and

determining a top border and a bottom border of the document bounding box by determining, from top to bottom of the image and from bottom to top of the image, respectively, whether a vertical histogram frequency value exceeds the predetermined threshold.

7. The method of claim 1 , wherein determining borders of the document bounding box includes determining whether the borders of the document bounding box are at least a minimum distance from corresponding edges of the image.

8. The method of claim 1 , further comprising:

determining a horizontal dispersion and a vertical dispersion of the horizontal and vertical histogram frequencies, respectively; and

resizing the document bounding box based on the determined horizontal and vertical dispersions.

9. The method of claim 8 , wherein resizing the document bounding box based on the determined horizontal and vertical centroids and the determined horizontal and vertical dispersions includes:

searching the horizontal histogram frequencies for a condition where an absolute value of a horizontal center of the document bounding box minus the horizontal centroid is no more than the product of a first predefined parameter and the horizontal dispersion; and

searching the vertical histogram frequencies for a condition where an absolute value of a vertical center of the document bounding box minus the vertical centroid is no more than the product of a second predefined parameter and the vertical dispersion.

10. A method of cropping an image based on image content, comprising:

receiving the image at a processor;

identifying image features of the image by determining a gradient of the image;

determining horizontal and vertical histogram frequencies of the identified image features;

estimating locations of borders of a document bounding box for cropping the image;

determining a horizontal centroid and a vertical centroid of the horizontal and vertical histogram frequencies, respectively;

refining the locations of the borders of the document bounding box based on the determined horizontal and vertical centroids; and

cropping the image to include the identified image features using the document bounding box and the refined locations of the borders.

11. The method of claim 10 , further comprising downscaling the image prior to the cropping the image.

12. The method of claim 11 , further comprising upscaling the document bounding box, wherein cropping the image using the document bounding box includes cropping the image using the upscaled document bounding box.

13. The method of claim 10 , wherein determining borders of the document bounding box includes examining the horizontal and vertical histogram frequencies of the identified image features.

14. The method of claim 13 , wherein examining the horizontal histogram frequencies of the identified image features includes determining a left border and a right border of the document bounding box, wherein the determined left and right borders contain image features having at least a predetermined feature count.

15. The method of claim 14 , wherein the left and the right borders have at least a minimum distance from corresponding left and right edges of the image, respectively.

16. A method of cropping an image, comprising:

receiving the image at a processor;

downscaling the image;

identifying image features of the downscaled image by determining a gradient of the image;

determining horizontal and vertical histogram frequencies of the identified image features;

examining the horizontal histogram frequencies of the identified image features to determine a left border and a right border of a document bounding box;

examining the vertical histogram frequencies of the identified image features to determine a top border and a bottom border of the document bounding box;

determining a horizontal centroid and a vertical centroid of the horizontal and vertical histogram frequencies, respectively;

refining a location of at least one of the left, right, top and bottom borders of the document bounding box if needed based on the determined horizontal and vertical centroids; and

cropping the image using the determined borders of the document bounding box.

17. The method of claim 16 , further comprising upscaling the document bounding box, wherein cropping the image using the determined borders of the document bounding box includes cropping the image using the upscaled document bounding box.

18. The method of claim 16 , wherein the left and the right borders and the top and the bottom borders have at least a minimum distance from corresponding edges of the image.

19. The method of claim 16 , wherein the left, right, top and bottom borders contain image features having a predetermined feature count.

Assignments (8)